Adaptive Agents Laboratory, Research Program


	RESEARCH PROGRAM	Agent Activities
		*Mavric*
		*Cyberbot*
		*Adaptrode*
		*Foraging Search*

Research

Publications

The exploration of intelligence by means of the synthetic approach is, by now, fairly well established as a discipline. The objective is to understand the nature of natural intelligence by building working models of intelligent agents. To the degree that these models successfully demonstrate intelligent behavior, we have a basis for analyzing the underlying mechanisms and drawing conclusions with some confidence.

Our approach is generally based on the concepts of Behavior-based Agents (also called Embodied Cognitive Science). These include:

Embodiment - the agent is not simulated
Situatedness - the agent exists in a real world
Composition of reactive atomic behaviors
Richness of sensory data
Cybernetic control concepts - feedback and feedforward
Adaptivity to a forever changing world

Adaptivity is the main focus of our efforts. Specifically we are interested in the question: How do animals adapt to changes which occur on multiple time scales relative to the life of the animal?

The main thrust of the lab at the present is the study of adaptive behavior in an embodied robot called MAVRIC. This robot learns to associate various cue events in its world (situated) with the availability of resources (food) or the presence of danger (poison). This research was funded in part by a grant from the National Science Foundation (IIS-9907102) and in part by a grant from the College of Arts & Sciences, Western Washington University.